2. Methodology

Linking the Weather Generator with Regional Climate Model: Effect of Higher Resolution

EGU2014-7344

Martin Dubrovsky1,3, Radan Huth1,2, Ales Farda3,4, Petr Skalak3,

Acknowledgements: The present experiment is made within the frame of projects ALARO-Climate (project P209/11/2405 sponsored by the Czech Science Foundation), WG4VALUE (project LD12029 sponsored by

the Ministry of Education, Youth and Sports of CR) and VALUE (COST ES 1102 action)

(1) Institute of Atmospheric Physics ASCR, Prague, Czech Rep.; (2) Dept. of Physical Geography and Geoecology, Faculty of Science, Charles University, Prague, Czech Rep.; (3) CzechGlobe - Center for Global Climate Change Impacts Studies, Brno, Czech Rep.; (4) Czech

Hydrometeorological Institute, Prague, Czech Rep.;

1. IntroductionLinking the weather generator (WG) with the Regional Climate Model (RCM) follows two purposes here: (1) Validation of RCM in terms of WG parameters, and (2) employing the RCM to assist the spatialization of WG (aiming to calibrate a gridded WG capable of producing realistic synthetic multivariate weather series for weather-ungauged locations).This contribution builds on our last year EGU poster, where the present-climate simulation by ALARO-Climate RCM (25 km resolution) was validated, and a methodology for linking the parametric WG with RCM output was introduced. Now we address a question: What is an effect of using a higher spatial resolution on a quality of simulating the surface weather characteristics? In the first part [BOX 1 + BOX 2], the new high resolution RCM simulation (the same model, but 6.25 km resolution) of the present climate will be validated in terms of selected WG parameters, which are derived from the RCM-simulated surface weather series and compared to those derived from weather series observed in 125 Czech meteorological stations. The set of WG parameters includes characteristics of the surface temperature and precipitation series. During the validation, we also demonstrate an effect of the higher resolution: the results obtained with 6.25 km resolution are compared with those obtained with the 25 km resolution.The second part [BOX 3] addresses a possibility of using the RCM (driven by reanalysis) to assist an interpolation of WG. In this case, the parametric M&Rfi WG, which is calibrated using the observed weather series from a set of stations, is spatialized(~interpolated) by fitting a spatial pattern of bias-corrected output from the Regional Climate Model.

List of WG parametersPwet …… probability of wet day occurrence (daily prec. amount ≥ 0.5 mm)PREC …… mean daily precipitation sum (= annual precipitation/365)Gs …… shape parameter of the Gamma distribution (used to approximate daily precip)Pdw ….... transitional probability of wet day occurrence (following the previous dry day)dPa …… mean amount of precipitation on a wet daya(T) …… long-term average of the daily mean temperature (T = TAVG)s(T) ……. standard deviation of TAVG deviations from its mean annual cyclelagcov(T) … lag-1(day) autocorrelation of TAVG

Czech Republic case study: RCM orography and altitudes of the 125 Czech weather stations

2.1 Regional Climate Model ALARO-Climate, which is presently being developed (within the frame of ALARO project) on the basis of the numerical weather prediction model ALADIN, is designed for simulations in common spatial resolutions (10 – 50 km) as well as in the "grey zone" resolutions (3 – 10 km). The ALARO physical parameterizations package includes microphysical processes parameterization 3MT (Modular Multiscale Microphysics and Transport), prognostic turbulence scheme (TOUCANS) and improved convection / radiation schemes. The present simulations were performed with ERA-Interim re-analysis for 25 and 6.25 km resolutions, ENSEMBLES domain, and 1961-1990 time slice.

2.2 Stochastic Weather Generator (M&Rfi) is a more flexible version of Met&Roll daily weather generator, which is a parametric WGEN-like generator: (i) precipitation occurrence ~ Markov chain model, (ii) precipitation amount ~ Gamma distribution, (iii) standardized non-precipitation variables ~ autoregressive model (first-order).

2.3 Validation of RCM in terms of WG parameters: see BOX 1 + BOX 2

2.4 RCM-assisted WG interpolation: see BOX 3

Z311

25 km

6 km

madu1110@gmail.com • this poster: www.ufa.cas.cz/dub/prasce/2014-egu-alaro~martin.pdf

• see poster Z364 (Wednesday) by Skalak et al. for more details on RCM

BOX 1: Validation of RCM in terms of 4 WG parameters ( 25 km vs. 6.25 km resolution)

[A] values of WG parameters [ Pwet, PREC, a(T), s(T)]: ● observed (weather stations)■ grid-specific RCM-based

[B] RCM vs OBS bias or ratio (RCM was interpolated to stations; OBS was “corrected” to the interpolated RCM altitude using observed vertical gradient)

- The mean biases and RCM-OBS correlations are shown in Box 2

grid = 25 km grid = 6.25 km grid = 25 km grid = 6.25 km

JanuaryJuly

25 km 6.25 km

Pwet

B (obs/RCM) B (obs/RCM) B (obs/RCM) B (obs/RCM)

PREC

B (obs/RCM) B (obs/RCM) B (obs/RCM) B (obs/RCM)

a(T)

B (obs-RCM)

A

B (obs-RCM)

A A A

B (obs-RCM) B (obs-RCM)

s(T)

B (obs/RCM) B (obs/RCM)B (obs/RCM)B (obs/RCM)

A

B

A AA A

A A A A

A AAA

BOX 3: RCM-assisted interpolation of the weather generator

RCM-assisted interpolation of 4 selected WG parameters (6.25 km RCM simulation is used):Gshape [summer] lagcov(TAVG) [summer]Pwet [winter] std(TAVG) [winter]

BOX 2: Validation of RCM in terms of 8 WG parameters – summary statistics

[A]

●OB

S ■

RC

Mra

w[B

] bi

ases

:O

BS

vs

RC

Mra

w[C

]●O

BS

■R

CM

corr

ecte

d

The procedure consists of following steps:1. determining WG parameters for RCM grids and 125 Czech weather stations → maps [A]2a. interpolating RCM-based WG parameters into stations and the reference altitude (~ 420 m)2b. “shifting” the OBS values into the reference altitude (using the mean observed vertical gradient)2c. determining RCM vs OBS biases (~ correction factors defined as differences or ratios) and interpolating them into RCM grids → maps [B]3. applying the interpolated correction factors to original RCM values → maps [C]

→ maps [A]

→ maps [B]→ maps [C]

The graphs show the mean bias (top) and correlation with OBS (bottom) for 8 WG parameters (4 of them are displayed in maps in Box 1) and 4 months.

The red and blue bars relate to 25 km and 6.25 km resolution, the dashed bars relate to a subset of stations, which lie above 500 m (36 of 125 stations).

• the results confirm that finer resolution improves the RCM vs. OBS fit (indicated by lower biases and higher correlations)

• the low [RCM, OBS] correlations mostly relate to WG parameters, which exhibit low spatial variability [e.g. s(T), lagcov(T), Gsh]

References: Regional Climate Model: ● Gerard L, Piriou J-M, Brozkova R, Geleyn J-F, Banciu D (2009) Cloud and precipitation parameterization in a meso-gamma scale operational weather prediction model. Mon. Wea. Rev. 137: 3960–3977. ● Benard P, Masek J, Smolikova P (2005) Stability of Leapfrog Constant-Coefficients Semi-Implicit Schemes for the Fully Elastic System of Euler Equations: Case with Orography. Mon. Wea. Rev. 133: 1065–1075. ● Farda A, Deque M, Somot S, Horanyi A, Spiridonov V, Toth H (2010) Model Aladin as regional climate model for Central and Eastern Europe. Studia Geoph Geodaetica 54: 313-332.Stochastic Weather Generator: ● Dubrovsky M, Zalud Z, Stastna M (2000) Sensitivity of CERES-Maize yields to statistical structure of daily weather series. Clim. Change 46: 447- 472. ● Dubrovsky M, Buchtele J, Zalud Z (2004) High-Frequency and Low-Frequency Variability in Stochastic Daily Weather Generator and Its Effect on Agricultural and Hydrologic Modelling. Clim. Change 63: 145-179. ● Dubrovsky M, Metelka L, Semeradova D, Trnka D, Halasova O, Ruzicka M, Nemesova I, Kliegrova S, Zalud Z (2006) The CaliM&Ro Project: Calibration of Met&Roll Weather Generator for sites without or with incomplete meteorological observations. in: Weather Types Classifications, Proc. 5th EMS Annual Meeting, Session AW8, O.E.Tveito and M.Pasqui, Eds. p.98-107.

B (obs/RCM)B (obs/RCM)

AA

CC

B (obs/RCM)

A

C

B (obs/RCM)

A

C